Document classification techniques for automated technology readiness level analysis

Abstract

Abstract The overhead of assessing technology readiness for deployment and investment purposes can be costly to both large and small businesses. Recent advances in the automatic interpretation of technology readiness levels (TRLs) of a given technology can substantially reduce the risk and associated cost of bringing these new technologies to market. Using vector‐space information‐retrieval models, such as latent semantic indexing, it is feasible to group similar technology descriptions by exploiting the latent structure of term usage within textual documents. Once the documents have been semantically clustered (or grouped), they can be classified based on the TRL scores of (known) nearest‐neighbor documents. Three automated (no human curation) strategies for assigning TRLs to documents are discussed with accuracies as high as 86% achieved for two‐class problems.